Lecture 3 Processor: Datapath and Control 1 ALU . Arithmetic Logic Unit is the hardware that pp,,,erforms addition, subtraction, AND, OR … 2 §4.1 Int r Recap: Performance oduction CPU Time Instruction Count CPIClock Cycle Time . CPU performance fftactors . Instruction count • Determined by Instruction Set Architecture and compiler . CPI and Cycle time • Determined by imppplementation of the processor Chapter 4 —The Processor — 3 Components of a Computer Processor . Datapath . Control . Components of the . Component of the processor that processor that perform arithmetic commands the operations and holds datapath, memory, data I/O devices according to the instructions of the memory §4.3 Bui Building a Datapath lding a D a . DhDatapath tapath . Elements that process data and addresses in the CPU • Memories, registers, ALUs, … . We will build a MIPS datapath incrementally . considering only a subset of instructions . To start, we will look at 3 elements Chapter 4 —The Processor —6 . A memory unit to store instructions of a program and supply instructi ons given an address . Needs to ppyrovide only read access (once the program is loaded). No control signal is need. 7 . PC (Program Counter or Instruction address register) is a register that holds the address of the current instruction . A new value is written to it every clock cycle. No control signa l is require d to enable write 8 . Adder to increment the PC to the address of the next instruction . An ALU ppyermanently wired to do onl y addition. No extra control signal required 9 Datapath portion for Instruction Fetch Increment by 4 for next 32‐bit instruction register Chapter 4 —The Processor — 10 Types of Elements in the Datapath . State element: . A memoryy,, element, i.e., it contains a state . E.g., program counter, instruction memory . Combinational element: . Elements that operate on values . EgE.g. adder, ALU 11 . Now, we will look at datapath elements reqqyuired by the different classes of instructions . Arithmetic and logical instructions . Data transfer instructions . Branch instructions 12 R-Format ALU Instructions . E.g., add $t1, $t2, $t3 . Perform arithmetic/logical operation . Read two register operands and write register result Chapter 4 —The Processor — 13 R-Format ALU Instructions . Register file: A collection of the registers . Any register can be read or written by specifying thhe number of the register . Contains the register state of the computer Chapter 4 —The Processor — 14 . Read from register file . 2 input s to the regist er file specifyi ng the numbers • 5 bit wide inputs for the 32 registers . 2 outputs from the register file with the read values • 32 bit wide . For all instructions. No control required. Chapter 4 —The Processor — 15 . Write to register file . 1 input to the register file specifying the number • 5 bit wide inputs for the 32 registers . 1 input to the register file wiihth the value to be written • 32 bit wide . OlOnly for some ins truc tions. RWitRegWritecontltrol silignal. Chapter 4 —The Processor — 16 . ALU . Takes two 32 bit input and produces a 32 bit output . Also, sets one-bit signal if the results is 0 . The operation done by ALU is controlled by a 4 bit control signal input. This is set according to the instruction Chapter 4 —The Processor — 17 Data transfer instructions . lw $t1, offset_value($t2) . Load: Read memory and update register . sw $t1, offset_value($t2) . Store: Write reggyister value to memory 18 Data transfer instructions . Compute the memory address by adding the value in base register and the 16 bit offset . need the ALU . Calculate address using 16-bit offset • Use ALU, but sign-extend offset . Write to or read from register . need the register file 19 . Two additional units – data memory and sign unit extension . Data memory . State element with • input for address and data to be written • output for read result . Data memory . Separate control for read and write . Control for read is required because reading from inva lid address can lead to problems . Sign-extension unit takes a 16 bit input and extend it to a 32 bit output 22 Composing the Elements for R-type and data transfer instructions . A simple data path that does an instruction in one clock cycle . Each datapath element can only do one function at a time . Hence, we need separate instruction and data memories . Use multiplexers where alternate data sources are used for different instructions Chapter 4 —The Processor — 23 Multiplexors . An ALU might need input from . Two registers . Or one registers and one immediate field (or off)ffset) . To choose correctly from multiple sources, a har dware element calle d multipl exor is used with appropriate control signals 24 Multiplexors . The data written to registers may come from . Data memory . Or ALU . To choose correctly from multiple sources, a hardware element called multiplexor is used withith appropriat e conttlrol siilgnals 25 R-Type/Load/Store Datapath Chapter 4 —The Processor — 26 Branch Instructions . beq $t1, $t2, offset . Read two registers and compare them . Take the 16 bit offset and add it to the address of next instruction following the branch ins truc tion to obta in the branc h targe t address Chapter 4 —The Processor — 27 Branch Instructions . Read register operands . Compare operands . Use ALU, subtract and check Zero output . Calculate target address . Sign-extend the offset . Shift left 2 places (word displacement) . Add to PC + 4 • Already calculated by instruction fetch Chapter 4 —The Processor — 28 Branch Instructions Just re‐routes wires Sign‐bit wire replicated Chapter 4 —The Processor — 29 Composing all elements together . Instruction fetch datapath . Datapath for R-type and memory instructions . Datapath for branches . Need an additional multiplexor to select the sequential address after branch or the branch tttarget address to be written to the PC 30 Datapath portion for Instruction Fetch Increment by 4 for next 32‐bit instruction register Chapter 4 —The Processor — 31 Full Datapath Chapter 4 —The Processor — 32 Datapath With Control AND gate for branch Chapter 4 —The Processor — 33 A Recap: Combinational Elements . AND-gate Adder A + Y . Y = A & B Y = A + B B A Y B Arithmetic/Logic Unit Multiplexer Y = F((,A, B) Y = S ? I1 : I0 A I0 M u Y ALU Y I1 x B S F Chapter 4 —The Processor — 34 A Recap: State Elements . Registers . Data Memory . Instruction Memory . Clocks are needed to decide when an element that contains state should be updated 35 Recap from Lecture 1: CPU Clocking . Operation of digital hardware governed by a constant-rate clock Clock period: duration of a clock cycle Clock frequency (rate): cycles per second 36 Clocks . A clock is a signal with a fixed cycle time (period) . The clock frequency is the inverse of the cycle time 37 Clocks . The clock cycle time or clock period is divided into two portions: . when the clock is high . whhen the cllkock is low 38 Clocking Methodology . We study . Edge triggered method ol ogy • Because it is simple . Edge triggered methodology: . All state changes occur on a clock edge Chapter 4 —The Processor — 39 Clocking Methodology : State Elements . Register: stores data in a circuit . Uses a clock signal to determine when to update the stored value . Edge-triggered: update when Clk changes from 0 to 1 Clk DQ D Clk Q Chapter 4 —The Processor — 40 Clocking Methodology : Stat e Element s . Register with write control . Only updates on clock edge when write control input is 1 . Used when stored value is required later Clk DQWrite Write D Clk Q Chapter 4 —The Processor — 41 Clocking Methodology . Combinational logic transforms data during clock cycles . Between clock edges . Input from state elements, output to state element • The state elements, whose outputs change only after the clock edggpe, provide valid inp uts to the combinational logic block. Chapter 4 —The Processor — 42 Clocking Methodology . To ensure that the values written into the state elements on the active clock edge are valid, the clock must have a long enough period so that all the signals in the combinational logic block stabilize, then the clock edge samples those values for storage in the state elements. This constraint sets a lower bound on the length of the clock period, which must be long enough for all state element inputs to be valid. Longest delay determines clock period Chapter 4 —The Processor — 43 It is possible to have a state element that is used as both an input and output to the same combinational logic block Ensure that the clock period is long enough 44 Single Clock Cycle . We studied a simple implementation where a singgyqle clock cycle is required for every instruction. Every instruction begins on one clock edge and completes execution on the next 45 Performance Issues . Longest dldelay determines cllkock period . Critical path: load instruction . Instruction memory register file ALU data memory register file . Not feasible to vary period for different instruct io ns . The clock cycle must be extended to accommodate the longest instruction . Improve performance by pipelining Chapter 4 —The Processor — 46 Conclusion . ISA influences the design of datapath and control for a processor . We studied an implementation bbdased on single cycle 47.